– Introduction
The ambition of the GPW13 is to improve the health of billions of people in the next five years. The WHO Impact Measurement Framework is part of GPW13. It measures progress at three levels:
- 46 outcome indicators and their global targets for 2023, covering a range of health issues
- The Triple Billion targets, to be achieved by 2023:
- billion more people benefiting from universal health coverage
- billion more people better protected from health emergencies
- billion more people enjoying better health and well-being.
- Healthy life expectancy (HALE), quantifying expected years of life in good health as a measure of the overall health of populations.
The purpose of this document is to be a live document that summarises the data “in hand” for the 46 outcome indicators i.e. the data that have been submitted to the DDI via the Global Health Observatory (GHO) or through other means. Based on either the estimates as generated by technical programmes or according to a set of estimation/extrapolation methodologies (with statistical assumptions and R-code provided), the document explores what estimates of baseline and projection values can be determined for each of the 46 indicators using the data we have and also highlights the indicators for which data gaps restrict robust estimation and extrapolation.
Detailed information on the methods of the GPW13 is available for download in the methods report which can be obtained from here,
– Table of indicators
The 46 outcome indicators are listed in the table below and can be downloaded from here. For each of these indicators, the ones highlighted in orange are included in the corresponding billions calculation whereas the ones highlighted in green are relevant to the billions but not directly included in the calculation.
Show Indicator Table Hide Indicator Table| GHO code | Indicator | UHC | HEP | HPOP | |
|---|---|---|---|---|---|
|
SDGDISASTER | Number of persons affected by disasters (per 100 000 population) | Relevant | ||
|
GHED_GGHE-DGGE_SHA2011 | Domestic general government health expenditure (GGHE-D) (% of general government expenditure (GGE)) | Relevant | ||
|
NUTRITION_HA_2 | Prevalence of stunting in children under 5 (%) | Included | ||
|
NUTRITION_WH_2 | Prevalence of wasting in children under 5 (%) | Included | ||
|
NUTRITION_WH2 | Prevalence of overweight in children under 5 (%) | Included | ||
|
MDG_0000000026 | Maternal mortality ratio (per 100 000 live births) | Relevant | ||
|
MDG_0000000025 | Proportion of births attended by skilled health personnel (%) | Relevant | ||
|
MDG_0000000007 | Under-five mortality rate (per 1000 live births) | Relevant | ||
|
WHOSIS_000003 | Neonatal mortality rate (per 1000 live births) | Relevant | ||
|
SDGHIV | New HIV infections (per 1000 uninfected population) | Relevant | ||
|
MDG_0000000020 | Tuberculosis incidence (per 100 000 population) | Relevant | ||
|
SDGMALARIA | Malaria incidence (per 1000 population at risk) | Relevant | ||
|
SDGHEPHBSAGPRV | Hepatitis B incidence (measured by: surface antigen (HBsAg) prevalence among children under 5 years) | Relevant | ||
|
SDGNTDTREATMENT | Number of people requiring interventions against NTDs | Relevant | ||
|
NCDMORT3070 | Probability of dying from any of CVD, cancer, diabetes, CRD (aged 30–70) (%) | Relevant | ||
|
SDGSUICIDE | Suicide mortality rate (per 100 000 population) | Included | ||
|
Coverage of treatment interventions for substance-use disorders (%) | Relevant | |||
|
SA_0000001688 | Total alcohol per capita consumption in adults aged 15+ (litres of pure alcohol) | Included | ||
|
RS_198 | Road traffic mortality rate (per 100 000 population) | Included | ||
|
SDGFPALL | Proportion of women (aged 15–49) having need for family planning satisfied with modern methods (%) | Included | ||
|
UHC_INDEX_REPORTED | UHC Service Coverage Index | Included | ||
|
FINPROTECTION_CATA_TOT_10_POP | Population with household expenditures on health >10% of total household expenditure or income (%) | Included | ||
|
SDGAIRBODA | Mortality rate attributed to air pollution (per 100 000 population) | Relevant | ||
|
SDGWSHBOD | Mortality rate attributed to exposure to unsafe WASH services (per 100 000 population) | Relevant | ||
|
SDGPOISON | Mortality rate from unintentional poisoning (per 100 000 population) | Relevant | ||
|
SDGTOBACCO | Prevalence of tobacco use in adults aged 15+ (%) | Included | Included | |
|
WHS4_100 | Proportion of population covered by all vaccines included in national programmes (DTP3, MCV2, PCV3) (%) | Included | Relevant | |
|
SDGHEALTHFACILITIESESSENTIALMEDS | Proportion of health facilities with essential medicines available and affordable on a sustainable basis (%) | Relevant | ||
|
HWF_0001 | Density of health workers (doctors; nurse and midwives; pharmacists; dentists per 10 000 population) | Included | ||
|
SDGIHR2018 | International Health Regulations (IHR) capacity and health emergency preparedness | Included | Included | |
|
Proportion of bloodstream infections due to antimicrobial resistant organisms (%) | Relevant | |||
|
SE_DEV_ONTRK | Proportion of children under 5 developmentally on track (health, learning and psychosocial well-being) (%) | Included | ||
|
SDGIPV | Proportion of women (aged 15–49) subjected to violence by current or former intimate partner (%) | Included | ||
|
SG_DMK_SRCR_FN_ZS | Proportion of women (aged 15–49) who make their own decisions regarding sexual relations, contraceptive use and reproductive health care (%) | Relevant | ||
|
WSH_WATER_SAFELY_MANAGED | Proportion of population using safely managed drinking water services (%) | Included | ||
|
WSH_SANITATION_SAFELY_MANAGED | Proportion of population using safely managed sanitation services and hand-washing facility (%) | Included | ||
|
SDGPOLLUTINGFUELS | Proportion of population with primary reliance on clean fuels (%) | Included | ||
|
SDGPM25 | Annual mean concentrations of fine particulate matter (PM2.5) in urban areas (µg/m3) | Included | ||
|
Proportion of children (aged 1–17) experiencing physical or psychological aggression (%) | Included | |||
|
Vaccine coverage for epidemic prone diseases | Included | |||
|
Proportion of vulnerable people in fragile settings provided with essential health services (%) | Included | |||
|
BP_04 | Prevalence of raised blood pressure in adults aged 18+ | Included | ||
|
NCD_CCS_SatFat | Effective policy/regulation for industrially produced transfatty acids (TFA) (Y/N) | Included | ||
|
NCD_BMI_30C | Prevalence of obesity (%) | Included | ||
|
WHS3_49 | Number of cases of poliomyelitis caused by wild poliovirus (WPV) | Relevant | ||
|
Patterns of antibiotic consumption at national level | Relevant | |||
| Note: | |||||
| a There are separate indicators for doctors, nurses, midwives and pharmacists in GHO, weighted average? | |||||
| b Sanitation and hand-washing have separate indicators in GHO, weighted average? | |||||
| c Adults and children have separate indicators in GHO, weighted average? | |||||
| * For the HPOP Billion non-age-standardized data are used. | |||||
| † SDG 1.a.2 includes several indicators. Only the one relating to health expenditure is used in GPW13. |
The 6 indicators without a gho_code in the table above are currently not in the GHO. The first do not have any data at all as yet nor do they have any additional info or focal points leading i.e.
- Coverage of treatment interventions for substance-use disorders (%) [UHC, SDG 3.5.1]
- Proportion of bloodstream infections due to antimicrobial resistant organisms (%) [UHC, SDG 3.d.2]
- Patterns of antibiotic consumption at national level [UHC, WHA 68.7]
The remaining three are components of the billions calculation and do not currently have data in GHO (need to follow up as to where they are and length of time-series) i.e.
- Proportion of children (aged 1–17) experiencing physical or psychological aggression (%) [HPOP, SDG 16.2.1] ⭐
- Vaccine coverage for epidemic prone diseases [HEP] ⭐
- Proportion of vulnerable people in fragile settings provided with essential health services (%) [HEP] ⭐
The remaining 40 indicators that are in the GHO can be extracted using the ghost R-package which was designed to provide a simple interface for extracting data from the World Health Organization’s Global Health Observatory (GHO) database using the GHO Open Data Protocol API. The package allows for exploration of indicators and dimensions available in the GHO and extract of these into R data frames.
– Methods for infilling and projections
Statistical models
Generalized Linear Models (GLMs)
A GLM has the structure \[ g\big(\mu_i\big) = \boldsymbol{x}_i \boldsymbol{\beta}, \] where \(\mu_i\) is the expectation of response variable \(Y_i\) such that \(\mu_i \equiv \mathbb{E}\big(Y_i\big)\), \(g\) is a link function, \(\boldsymbol{x}_i\) is the \(i\)th row of the covariate matrix \(\boldsymbol{X}\), and \(\boldsymbol{\beta}\) is a vector of unknown coefficients. Akin to linear models, GLMs center around a linear predictor, \(\boldsymbol{X}\boldsymbol{\beta}\), and allow for link functions other than the identity function. GLMs also allow for any distribution from the exponential family to model the dependent variable instead of just assuming it Gaussian. The exponential family consists of distributions whose probability density function can be written as \[ f_{\theta}\big(y\big) = \exp \bigg( \frac{y \theta - b(\theta)}{a(\Phi)} + c(y, \Phi)\bigg), \] where \(a\), \(b\), and \(c\) are arbitrary functions, \(\Phi\) is a scale parameter, and \(\theta\) is the canonical parameter of the distribution.
For the projections or infilling univariate models by country and indicator are used.
Generalized Additive Models (GAMS)
Building upon GLMs, a GAM has the structure:
\[ g\big(\mu_i\big) = \boldsymbol{x}_i^* \boldsymbol{\theta} + f_{1}\big(x_{1, i}\big) + f_{2}\big(x_{2, i}\big) + f_{3}\big(x_{3, i}, x_{4, i}\big) + \ldots, \]
where \(\mu_i\) again is the expectation of response variable \(Y_i\), i.e., \(\mu_i \equiv \mathbb{E}\big(Y_i\big)\), the response variable \(Y_i\) follows any distribution from the exponential family, parametric model components have model matrices with rows \(\boldsymbol{x}_i^*\), the vector \(\boldsymbol{\theta}\) is of the corresponding parameters, and covariates \(x_k\) are smoothed with the functions \(f_j\).
The smooth functions are non-parametric estimates of the actual functional relationship between the response variable and the predictor variable to which the smooth function is applied. Unlike LMs or GLMs, a smooth function is non-parametric, i.e., it does not assume the response variable to follow any specific distribution. As an explanatory example, consider a model containing one smooth function of one predictor variable with the identity function as the link function,
\[ y_i = f\big(x_i\big) + \varepsilon_i, \]
where \(y_i\) is a response variable, \(x_i\) is a predictor variable, \(f\) is a smooth function, and the \(\varepsilon_i\) follow the normal distribution with a mean of zero and a variance of \(\sigma^2\). GAMS can also be used for time-series modelling with the form \[y_t = f(log(t)) + \varepsilon_{t}\] where \(y_t\) is the time-series variable to be forecasted using time \(t\). The smoothing function can be on \(t\) or on \(log(t)\) to determine moderate forecasts. The model can be assumed to have autoregressive lag-1 (AR-1) errors:
\[\varepsilon_{t} =\psi_1 \varepsilon_{t-1} + w_t\] where \(w_t \sim \text{ iid }N(0,\sigma^2)\). The \(\psi_1\) correlation is estimated and accounted for in the parameter estimation for the smoother \(f(\cdot)\).
For the projections or infilling univariate models by country and indicator are used.
GBoostM
GBoostM is shorthand for Gradient Boosting Machines. GBoostM is one of the most powerful learning ideas introduced in the last twenty years. It is a machine learning technique for regression and classification problems which produces a prediction model in the form of an ensemble of weak prediction models, building the final model in an iterative fashion.
The purpose of boosting is to sequentially apply the weak classification algorithm to repeatedly modified versions of the data, thereby producing a sequence of weak classifiers \(G_m(x)\), \(m = 1, 2, ... , M\) that together reduce the error in the prediction.
The final model finds the correct mathematical manipulations to turn the input into the output. For the projection and infilling, the GBoostM works on the entire pooled indicator dataset across all countries.
Fixed logit
Booth conducted a review of forecasting methods related to health and specifically mortality and grouped them into three broad categories. The first group comprises of the extrapolation methods which make use of the regularity of trends over time. A second group is made up of the explanation methods which utilize structural or epidemiological models and covariates in the causal pathway to forecast. A third group can be defined as expectation, for which forecasts are based on the opinions of experts. The result is generally a fixed rate of increase or decrease for the forecasted measure based on the information available to the expert. The challenge with this approach is the subjectivity of either the direction or magnitude of the change moving forward, which may not have reproducible reasoning to support validity.
A crude assumption can be made to project indicators under the expectation umbrella by assuming that the rate of change in the indicators for the next few years is very similar to the average logit change in the most recent years and that this change is not expected to persist with the same intensity but rather to diminish with time.
Average
The series labelled Average gives the final projection estimate. It takes the average of the various statistical models used for the particular country-year-indicator combination. If only a GAM model is used, for example, then the Average is equivalent to the mean GAM whereas if both GLM and GAM are used, then the Average is calculated using the means of the GAM and GLM series. In general the simple heuristic applied for the projection is that if there is only a single data point for the indicator-country, then we turn to the GBoostM to complete the series. As the boosting pools all the data it is sometimes less accurate in-sample, as compared to the GAM, GLM and Fixed logit difference - which work specifically on a single country time-series. If there are at least two data points, we can calculate the Fixed difference, for greater than five data points we can also determine the GLM and for ten or more points we can generate the GAM. Depending on how many data points we have and which series are included, we can then calculate this Average.
Demographic Index covariate
Baselines and projections for the 40 GHO indicators as well as the additional 9 UHC are either determined by the technical programmes or by the DDI. To build a systematic approach to infilling and extrapolating estimates for all countries we can use time as the main predictor as well as other social and demographic indices. A composite covariate developed for this report is a version of the Socio-demographic Index (SDI) that was originally developed for the 2015 Global Burden of Disease study. This development status indicator has components that are strongly correlated with health outcomes. It is the geometric mean of normalised 0 to 1 indices of total fertility rate, mean years of schooling for those aged 15 and older, and lag distributed income per capita. Instead of using the GBD2017 SDI (some of the input data for this measure are not publically available e.g. fertility), we construct the metric using publicly available fertility data from WPP2019 as well as GDP per capita and education data from the UN Human Development Reports. In the plot below, we compare this constructed SDI to the one from the GBD.
In general, there is strong agreement, between the SDI measure from the public data vs the one from the GBD. There are some missing values for a few countries in the public data and these can be imputed using a spline log-linear model that predicts the sdi.pub using the sdi.gbd and time. A summary of the regression results for such a model is shown below.
| sdi.pub | |||
|---|---|---|---|
| Predictors | Estimates | CI | p |
| (Intercept) | 0.54 | 0.54 – 0.54 | <0.001 |
|
region [Europe & Central Asia] |
-0.00 | -0.01 – 0.00 | 0.888 |
|
region [Latin America & Caribbean] |
0.05 | 0.05 – 0.06 | <0.001 |
|
region [Middle East & North Africa] |
-0.02 | -0.03 – -0.01 | <0.001 |
| region [North America] | 0.03 | 0.02 – 0.05 | <0.001 |
| region [South Asia] | -0.01 | -0.02 – -0.00 | 0.010 |
|
region [Sub-Saharan Africa] |
-0.00 | -0.01 – 0.00 | 0.199 |
| Smooth term (sdi.gbd) | 8.29 | <0.001 | |
| Smooth term (lt) | 1.42 | 0.218 | |
| Observations | 3140 | ||
| R2 | 0.968 | ||
Forecasting the SDI covariate
One of the main advantages of a covariate model is that it can be used to infill missing values in the retrospective data and once projected to the target year, it can be used to increase the robustness of any forecasts of the indicators. The SDI estimates for each country are forecasted using GAMS. Plots of the country and region specific SDI estimates and forecasts can be seen below.
Show SDI forecasts Hide SDI forecasts
– Summary assessment of indicators
| SDG/WHA indicator no | Indicator | Family | Summary | Action |
|---|---|---|---|---|
|
Number of persons affected by disasters (per 100 000 population) | HEP | All countries but only has data for 2016 | Update with GHE for baseline and projections (if any) |
|
Domestic general government health expenditure (GGHE-D) (% of general government expenditure (GGE)) | UHC | Robust time series to 2017 for almost all countries | Can generate baselines/projections (or tech prog if available), country consultation if we generate |
|
Prevalence of stunting in children under 5 (%) | HPOP | Scattered data points, some countries with baselines | To rely on tech progs and if they cannot supply, can estimate, country consultation required for some |
|
Prevalence of wasting in children under 5 (%) | HPOP | Scattered data points, some countries with baselines | To rely on tech progs and if they cannot supply, can estimate, country consultation required for some |
|
Prevalence of overweight in children under 5 (%) | HPOP | Scattered data points, some countries with baselines | To rely on tech progs and if they cannot supply, can estimate, country consultation required for some |
|
Maternal mortality ratio (per 100 000 live births) | UHC | Robust time series to 2017 for almost all countries | Update with GHE for baseline and projections |
|
Proportion of births attended by skilled health personnel (%) | UHC | Scattered data points, some countries with baselines | Can generate baselines/projections (or tech prog if available), country consultation if we generate |
|
Under-five mortality rate (per 1000 live births) | UHC | Robust time series to 2018 for all countries | Update with GHE for projections |
|
Neonatal mortality rate (per 1000 live births) | UHC | Robust time series to 2018 for all countries | Update with GHE for projections |
|
New HIV infections (per 1000 uninfected population) | UHC | Robust time series to 2019 for many countries (not 194 in database) | To rely on tech programs for projections |
|
Tuberculosis incidence (per 100 000 population) | UHC | Robust time series to 2018 for all countries | To rely on tech programs for projections |
|
Malaria incidence (per 1000 population at risk) | UHC | Robust time series to 2018 for relevant countries | To rely on tech programs for projections |
|
Hepatitis B incidence (measured by: surface antigen (HBsAg) prevalence among children under 5 years) | UHC | All countries but only has data for 2015 | Check tech program otherwise maybe look at Hep mortality as covariate, need to derive baselines and projections |
|
Number of people requiring interventions against NTDs | UHC | Robust time series to 2018 for all countries | To rely on tech programs for projections?? |
|
Probability of dying from any of CVD, cancer, diabetes, CRD (aged 30–70) (%) | UHC | Scattered data points in GHO but GHE will have time series and baselines | Update with GHE for projections and baselines |
|
Suicide mortality rate (per 100 000 population) | HPOP | Scattered data points in GHO but GHE will have time series and baselines | Update with GHE for projections and baselines |
|
Coverage of treatment interventions for substance-use disorders (%) | UHC | No data | Ignore?? |
|
Total alcohol per capita consumption in adults aged 15+ (litres of pure alcohol) | HPOP | Scattered data points, some countries with baselines | To rely on tech progs and if they cannot supply, can estimate, country consultation required for some |
|
Road traffic mortality rate (per 100 000 population) | HPOP | All countries but only has data for 2016 | Update with GHE for baseline and projections |
|
Proportion of women (aged 15–49) having need for family planning satisfied with modern methods (%) | UHC | Scattered data points, some countries with baselines | To rely on tech progs and if they cannot supply, can estimate, country consultation required for some |
|
UHC Service Coverage Index | UHC | Composite of other indicators, generally using GLM or GAM for extrapolation | Can generate baselines/projections (or tech prog if available), country consultation if we generate |
|
Population with household expenditures on health >10% of total household expenditure or income (%) | UHC | Scattered data points, some countries with baselines | To rely on tech progs and if they cannot supply, can estimate, country consultation required for some |
|
Mortality rate attributed to air pollution (per 100 000 population) | HPOP | All countries but only has data for 2016 | Update with GHE for baseline and projections |
|
Mortality rate attributed to exposure to unsafe WASH services (per 100 000 population) | UHC | All countries but only has data for 2016 | Update with GHE for baseline and projections |
|
Mortality rate from unintentional poisoning (per 100 000 population) | HPOP | All countries but only has data for 2016 | Update with GHE for baseline and projections |
|
Prevalence of tobacco use in adults aged 15+ (%) | UHC & HPOP | Not all countries but all in data base have 2018 | To rely on tech progs and if they cannot supply, can estimate projection |
|
Proportion of population covered by all vaccines included in national programmes (DTP3, MCV2, PCV3) (%) | UHC & HEP | Robust time series to 2018 for all countries | To rely on tech progs and if they cannot supply, can estimate projection |
|
Proportion of health facilities with essential medicines available and affordable on a sustainable basis (%) | UHC | Very small number of observations (countries and years) | Can possibly model? But utility of indicator for 25 countries? |
|
Density of health workers (doctors; nurse and midwives; pharmacists; dentists per 10 000 population) | UHC | Robust time series to 2018 for most countries | To rely on tech programs for projections and baselines for all countries |
|
International Health Regulations (IHR) capacity and health emergency preparedness | HEP | Almost all countries with data for 2018 and 2019 | Will have to determine projections |
|
Proportion of bloodstream infections due to antimicrobial resistant organisms (%) | UHC | No data | Ignore?? |
|
Proportion of children under 5 developmentally on track (health, learning and psychosocial well-being) (%) | HPOP | Only subset of countries and subset of years, some max as far back as before 2010 | Needs more data, was this from technical program? Need to find more data or to model |
|
Proportion of women (aged 15–49) subjected to violence by current or former intimate partner (%) | HPOP | Only subset of countries and subset of years, some max as far back as before 2011 | Needs more data, was this from technical program? Need to find more data or to model |
|
Proportion of women (aged 15–49) who make their own decisions regarding sexual relations, contraceptive use and reproductive health care (%) | HPOP | Only subset of countries and subset of years, some max as far back as before 2012 | Needs more data, was this from technical program? Need to find more data or to model |
|
Proportion of population using safely managed drinking water services (%) | HPOP | Robust time series to 2017 for about 90 countries | Can generate baselines/projections (or tech prog if available), country consultation if we generate |
|
Proportion of population using safely managed sanitation services and hand-washing facility (%) | HPOP | Robust time series to 2017 for about 90 countries | Can generate baselines/projections (or tech prog if available), country consultation if we generate |
|
Proportion of population with primary reliance on clean fuels (%) | HPOP | Robust time series to 2018 for most countries | To rely on tech progs and if they cannot supply, can estimate projection |
|
Annual mean concentrations of fine particulate matter (PM2.5) in urban areas (µg/m3) | HPOP | All countries but only has data for 2016 | There is a model but no known projections, need to discuss with program. May have to level extrapolate. |
|
Proportion of children (aged 1–17) experiencing physical or psychological aggression (%) | HPOP | Need to get SDG data | Data will be cross sectional so need to decide on rule of thumb for projection |
|
Vaccine coverage for epidemic prone diseases | HEP | No data | Ignore?? |
|
Proportion of vulnerable people in fragile settings provided with essential health services (%) | HEP | No data | Ignore?? |
|
Prevalence of raised blood pressure in adults aged 18+ | UHC | Robust time series to 2015 for most countries | To rely on tech programs for projections and baselines for all countries |
|
Effective policy/regulation for industrially produced transfatty acids (TFA) (Y/N) | HPOP | Table for 2019 | Assume 2019 values |
|
Prevalence of obesity (%) | HPOP | Robust time series to 2016 for all countries | Can generate baselines/projections (or tech prog if available), country consultation if we generate |
|
Number of cases of poliomyelitis caused by wild poliovirus (WPV) | HEP | Number of countries with baselines | Could just be assumed 0 I am sure, having been "“eliminated”" |
|
Patterns of antibiotic consumption at national level | UHC | No data | Ignore?? |
| UHC 1 | Family planning | UHC | Robust time series to 2019 for almost all countries | GLM works well if technical program does not project |
| UHC 2 | Suspected pneumo | UHC | Robust time series to 2016 forabout 120 countries | A covariate model on U5MR exists and is supposed to work well. Can compare and rather use that. Consultation required for baselines |
| UHC 3 | Population at risk of Malaria who slept under treated net | UHC | Pretty noisy but has baseline for endemic countries | Potentially use average of linear and GAM for projection |
| UHC 4 | Proportion of population using basic sanitation | UHC | Robust time series to 2016/17 for all countries | GLM works well if technical program does not project |
| UHC 5 | Proportion of TB cases that are treated | UHC | Robust time series to 2016/17 for all countries | GLM works well if technical program does not project |
| UHC 6 | Proportion of people living with HIV on ART | UHC | No all countries but necessary ones have baselines | Will use tech programs baselines and projections for relevant countries |
| UHC 7 | Mean fasting plasma glucose | UHC | For all countries and data up to 2014. Need to reassess utility of this indicator and if other can be found | GLM works well enough but issue of country consultation of baselines |
| UHC 8 | Number of Hospital beds available per 10,000 | UHC | Some countries have data to 2018 others do not | Can use average or hierarchy of models (Fixed > Linear > GAM > GBoostM). Will require consultation for some |
| UHC 9 | Proportion of wmen receiving antenatal care | UHC | Mixture, some have baselines and some do not | Can use average or hierarchy of models (Fixed > Linear > GAM > GBoostM). Will require consultation for some |
– Download links for databases
- The data extracted from the GHO can be downloaded as an R-database using the GHO_data.RData link below:
- All of the input data used in the models and to generate summary tables can be downloaded as an R-database using the Input_data.RData link below:
- All of the output data generated by the models can be downloaded as an R-database using the Output_data.RData link below: